provider crawls all tables and excludes rather than just including specified tables

[MikeG]

If you have:

[mysqldump]
tables = "foo.importantTable'

You will see holland do something along the lines of:

Excluding mysql.columns_priv
Excluding mysql.db
Excluding mysql.func
Writing ignore-table exlusions to /var/spool/holland/bkset/20101120_022118/my.cnf

This could be really inefficient on a server with thousands of tables when you want just the one

[Andrew Garner (abg)]

Basically what holland-mysqldump does is, given a set of inclusion/exclusion filters generate a mysqldump command line to backup the requested tables. However, mysqldump itself has *very* primitive support for inclusion/exclusions. This support basically falls into three categories:

1) Dump specific tables from *one* database only => mysqldump [options] database table [table..]

2) Dump multiple databases and exclude specific tables => mysqldump [options] --ignore-table=db.tbl --databases db [db...]

3) Dump all databases and exclude specific tables => mysqldump [options] [--ignore-table=db.tbl...] --all-databases

If we have a single 'tables' request, we may be able to hard code a fast path for that to always choose (1). If it includes a pattern we still have to scan a potentially large number of tables to find matches. If "tables" includes tables from multiple databases but file-per-database=off (meaning, the user wants a consistent snapshot of all those tables via a single mysqldump command) there's not a good way to support that without generating a large list of --ignore-table options.

I think the common use case is to simply filter a few tables and otherwise back everything else up. You can always manipulate the mysqldump command line through holland via additional-options. For efficient inclusion/exclusion we would need support from mysqldump (or a similar backend). I'm skeptical that the additional complexity in holland-mysqldump to more efficiently handle edge cases here is warranted currently. Although perhaps I'm just not aware of interesting use cases.

Essentially I view this as a limitation in mysqldump itself and holland has a reasonable general purpose filtering algorithm. I'm completely open to contributions to provide better alternatives - either as new plugins or reworking the existing "Given a set of filters, run these mysqldump commands" algorithm.